Replies to post #197670 on Biotech Values

[iwfal]

BMRN -

the one thing I find a bit harder to sign off on is several families claiming their son would deteriorate during the period drug was suspended and then stabilize after drug was started (sometime afterward). From what I've learned in DMD that is unusual.

At least for 6MWD it isn't uncommon to drift up and down some - see, for instance, Goemanns. What would be odd is a bunch of patients stabilizing at, say, 200m for 2 years.

[iwfal]

BMRN FDA Review and "Easier FDA" (a merging of two threads)

I would suggest that the BMRN FDA review (particularly in light of the questions) suggests that I was correct in #msg-116782774 . They are probably easier only around the margins, only in one way and to a significant degree that 'new easiness' is more about stronger therapies than really getting easier.

the FDA does seem to have become somewhat less stringent about strict statistical rules for efficacy, but for many of the 'extra' drugs there is no appreciable risk that they have little efficacy. I.e. at some point, with enough efficacy (e.g. HR<0.2 and p<0.00001 vs a generally accepted clinical endpoint like PFS for AA), it is a virtual certainty the drug is useful even without a good pre-specified protocol. In fact a reasonable portion of the ‘extra’ approvals looks like Break Thru Designation going all the way to approval.

I.e. there are multiple ways that the FDA could get 'easier'. Several of the more obvious are below:

a) get a little less retentive about 'pre-specified' endpoints as long as the endpoints were historically reasonable and the p value stellar.

b) look more at "Totality of efficacy data". E.g. if a drug misses on all 3 of its RCTs with ITT p values on the pre-specified endpoints with, say, p=0.1 on one, p=0.15 on another and p=0.2 on the last then approve it because in aggregate (before baserate correction) the p value is 0.000375 (much less than the oft quoted 0.000625 one sided).

c) allow more patient discretion wrt SAE risk. I.e. if the trials were big enough to show the risk then risk is not an approval factor because it is up to the patients to make the choice.

d) less proof that biomarkers predict clinical disease.

I suggested that #a was the primary place that the FDA has gotten easier and the BMRN AdComm appears to confirm that it was NOT getting easier in #b-#d. (I underlined 'appears' because until the FDA rules it is still preliminary)

Random other comments:

1) that I think the RCT data for Dris does prove efficacy fairly strongly (see #b above), but do agree that the single arm data is hopelessly worthless without *MUCH* better historical data (this is a change from my previous views). Note that that bodes extremely poorly for SRPT.

2) Further comments on #b above - that although it is not a new phenomenon, it is somewhat of a mystery why the FDA doesn't look at 'totality of (ITT RCT) data'. My *guess* is that they give too much credence to the 'sensitivity' analyses (e.g. what happens to p value is patient x were different) and those are inherently more damaging when the p value is spread over many different ITT RCTs. (Note: a corollary to my oft stated claim to be able to post hoc any wiffed trial into stat sig is that I can equally easily destroy any trial with p>0.01)

3) Reminder - there is a fair amount of arbitrariness to ultimate FDA approval. E.g. there are generally several what-the-heck approvals every year. And many times the FDA appears to CRL just to get the company to respond with some quick fix data. But regardless the arbitrariness means that the above commentary is, at best, subject to noise.